Abstract
Newberry Volcano in Oregon, USA, has two small crater lakes inside its caldera: East Lake and Paulina Lake. The 50–80 m deep lakes differ in morphology, water chemistry and sediment composition, although separated only by a narrow volcanic ridge. East Lake is a terminal lake with gaseous geothermal inputs, whereas Paulina Lake has an outlet (Paulina Creek) and subaqueous, high-alkalinity hot springs. The sediment in both lakes is organic-rich and mainly consists of diatom frustules (SiO2) with some volcanic ash, accumulating at sedimentation rates of 1.5–2 mm a−1 (210Pb dated). In Paulina Lake the sediment has up to 14% Fe2O3 of hydrothermal origin and 250 ppm As. Sediment in East Lake is Fe-poor but has up to 4 ppm Hg, and fish are also Hg-rich. Both lakes host productive ecosystems, with primary producers using geothermal CO2, P and Si, and nitrogen fixed by cyanobacteria (Nostoc sp.). Water budgets and water residence times were calculated from stable isotope budgets. East Lake has a steep vertical δ13C (dissolved inorganic carbon, DIC) gradient, with surface δ13C values of up to 5.5‰, largely due to diffusional CO2 losses and photosynthetic carbon withdrawal. Paulina Lake is better mixed, has a lower organic productivity and limited surface CO2 evasion; its δ13C (DIC) gradient is small. We theorize that the lakes are fed by different geothermal components as a result of phase separation below East Lake. The gas component (CO2, H2S, Hg) enters East Lake, as indicated by the rising bubble trains in the lake. A residual fluid, depleted in Hg but rich in Si, Fe, carbonate, P and As, enters Paulina Lake. The presence of highly toxic components and the gas-charging of East Lake present natural hazards, which may change when new volcanic activity is initiated, and thus should be monitored.
Volcanic lakes in dormant volcanoes can be affected by volcanic and geothermal effluents and thus become polluted by volcanogenic components such as F, Cl or SO4, or by toxic trace elements such as B, Li, As or Hg (Löhr et al. 2005; Varekamp 2015). These volcanic signatures may be apparent in the water column and sediment (Nelson et al. 1994; Lottermoser et al. 1997; Miyabuchi & Terada 2009). Volcanic lake ecosystems may also be influenced by volcanic inputs, such as P and gaseous CO2. The latter may lead to hazardous accumulations of CO2 in meromictic lakes without regular mixing of the hypolimnion, e.g. Lake Nyos, Cameroon (Kling et al. 1989; Zhang & Kling 2006; Tassi & Rouwet 2014). The changing composition of volcanic lake waters can be used to monitor the host volcano as well as its limnic system (e.g. Pasternack & Varekamp 1997; Rouwet & Tassi 2011; Varekamp 2015).
We present comprehensive data from two crater lakes in the Newberry Volcano caldera in central Oregon, USA. These two closely spaced lakes are famous local fishing grounds, although the fish in one lake have high Hg levels (Newell et al. 1996; Stone et al. 1996). Hot springs dot the beaches (Russell 1905; Forcella 1982; Johnson 1985; Ingebritsen et al. 2014), reflecting geologically recent volcanic activity, with the last eruption c. 1.3 ka ago (Sherrod et al. 1997). We provide several years of limnological data, geological observations, chemical profiles and C–H–O stable isotope values for the water column and sediment C–N isotope data. We provide a water budget that is based on stable isotope modelling and a carbon cycle model for both lakes that is based on δ13C (dissolved inorganic carbon, DIC) gradients.
Newberry Volcano
Newberry Volcano (43° 41.21′ N, 121° 15.18′ W) is 65 km east of the Cascade Range and 30 km south of Bend, Oregon (Fig. 1). Despite rising only c. 1200 m above the surrounding High Lava Plains, Newberry Volcano is one of the largest Quaternary volcanoes in the conterminous USA (Russell 1905; Higgins 1973; MacLeod & Sherrod 1988). Its flanks slope gently upwards to the 8 km diameter caldera (Higgins 1973), which formed c. 75 ka BP and is partly filled with Pleistocene and Holocene pyroclastics, flows, domes and lacustrine sediments.
Location of Newberry Volcano with its twin crater lakes.
The two, small, 50–80 m deep crater lakes in the caldera (Figs 1 & 2), East Lake (EL) and Paulina Lake (PL) (Johnson 1985), form the focus of this study. Conifer trees, including lodge-pole pine, yellow pine, alders and firs, cover most of the caldera (Russell 1905; Johnson 1985). The abundant basaltic cinder cones and fissure vents are generally aligned NNE, parallel to the Walker fault zone, but on the north flank they are aligned NNW, parallel to the Sisters fault zone. The upper portions of the flanks are composed of dacite, rhyodacite and rhyolite domes and flows (Fig. 3). Holocene basaltic andesite flows are found on the flanks and Holocene rhyolitic material occurs in the caldera (Linneman 1990; MacLeod et al. 1995). The erupted material has been grouped into six volcanic stages on the basis of chemical composition and stratigraphic relations to the Mazama ash layer (MacLeod & Sammel 1982).
The Newberry caldera, looking NE from Paulina Peak. PL is on the left, and EL in the top right. The non-vegetated area in the bottom right is the Big Obsidian Flow. The Central Pumice Cone is visible between the two lakes.
Geological map of the Newberry caldera. Holocene silicic material is the dominant rock type within the caldera and basalt flows cover the flanks of the volcano (from MacLeod & Sammel 1982).
Units below the Mazama ash have an estimated age of 8–10 ka, (MacLeod & Sammel 1982; MacLeod & Sherrod 1988), and include rhyolitic domes and obsidian followed by mafic flows and scoria (MacLeod & Sherrod 1988). Silicic volcanics above the Mazama ash were dated c. 6.2 ka BP (MacLeod & Sherrod 1988) and are overlain by basaltic andesite with a 14C age of 6.1 ka. A phreatomagmatic eruption, probably from a vent below the current EL, deposited thick pumiceous tephra in the caldera as well as smaller pumice rings and obsidian intrusions in the southern and eastern parts of the caldera. The narrow ridge between the two lakes consists of the Central Pumice Cone and the Interlake Obsidian Flow, both of which formed during this same period. Younger fissures erupted obsidian flows and pumices with an estimated age of c. 3.5 ka BP (MacLeod & Sherrod 1988). The most recent major volcanic event produced air-fall tephra, ash-flows and obsidian. The PL ash flow and the Big Obsidian Flow were probably erupted from the same vent on the southern wall of the caldera and extend north, the latter covering about 20 km2 of the caldera. Radiocarbon ages suggest emplacement around 1.3–1.35 ka BP. Jensen & Chitwood (2000) documented ongoing caldera floor uplift from the examination of submerged and elevated PL shoreline terraces. Evidence for a poorly dated catastrophic flood down Paulina Creek (PCR) was provided by Chitwood & Jensen (2000).
The Newberry geothermal system
Geothermal energy exploration provided data on the composition of the geothermal fluids and a 932 m long USGS drill core (Newberry 2) displays varied hydrothermally altered lithologies (Keith & Bargar 1988; Swanberg et al. 1988). At depths of 290 m, a 30 m thick bed of hydrated basaltic sand, siltstone and mudstone with graded beds, cross-beds and flame structures interrupts the volcanic flows and clastic debris. This bed has been interpreted as fluvio/lacustrine, suggesting the presence of a large caldera lake (Keith & Bargar 1988). Above the lacustrine unit basaltic and rhyolitic lapilli, tuff, breccia and alluvium are found. Glassy fragments are common, suggesting subaqueous eruptions. Immediately below the lacustrine unit are rhyodacite pumices and tuffs, interpreted as ash flows, possibly related to caldera collapse events (Higgins 1973; Keith & Bargar 1988).
The core of Newberry Volcano is probably a complex cluster of cooling magmatic intrusions (Stauber et al. 1988), with a thermal chimney that may focus the ascent of new magma and thermal energy (Fitterman 1988; MacLeod & Sherrod 1988). The Newberry 2 drill core has a maximum recorded temperature of 265°C (MacLeod & Sammel 1982). Fluids are probably transported along fractures, faults and brecciated intrusions (MacLeod & Sammel 1982). Fluids from the bottom 2 m of the drill hole consist primarily of steam and CO2 (Priest et al. 1983; Ingebritsen et al. 1986). Geophysical studies all suggest the presence of magmatic stocks at 1–3 km depth below the caldera floor (Achauer et al. 1988; Catchings & Mooney 1988; Gettings & Griscom 1988; Stauber et al. 1988). Hot springs and fumaroles in the caldera derive from this active hydrothermal system, and suggest the possibly continued existence of hot magma at depth (Forcella 1982; Johnson 1985; Ingebritsen et al. 2014).
Descriptive limnology and caldera hydrology
The two lakes are separated by a c. 2 km wide ridge of Holocene volcanics. Both EL and PL have bubbling hot springs along their peripheries, indicating input from subsurface hydrothermal fluids (Russell 1905; Phillips & Van den Burgh 1968; Forcella 1982; Johnson 1985; Lefkowitz 2012). EL, with a surface elevation of c. 1945 m above mean sea-level (amsl), is the smaller and shallower lake, with a surface area of 4.2 km2, a maximum depth of 55 m, an average depth of 20 m and a volume of 86×106 m3 (Johnson 1985). About 50% of its surface area has a water depth <11 m (Figs 4 & 5). PL, with a surface elevation of c. 1930 m amsl has a surface area of 6.2 km2, a maximum depth of 76 m, an average depth of 50 m and a volume of 310×106 m3 (Johnson 1985). About 50% of its surface area has a water depth of more than 60 m (Figs 4 & 5). Sammel & Craig (1983) suggested the occurrence of seepage down the hydraulic gradient from El to PL, although geochemical evidence for such a process is lacking.
Comparison of the lake morphologies of PL and EL.
Bathymetric maps of (a) PL (contours are in metres) and (b) EL.
EL and PL have steeply sloping drainage basins of comparable size (22 and 21 km2 respectively; Johnson 1985). EL is a terminal lake, whereas PL has an overflow through PCR (Russell 1905; Johnson 1985; Morgan et al. 1997; Sherrod et al. 1997), which breaches the western caldera wall to discharge into the Little Deschutes River valley c. 24 km west of PL.
The regional water table under the Newberry Volcano is estimated at c. 1275 m amsl, and the caldera probably contains multiple perched aquifers within permeable hydrogeological units, underlain by impermeable volcanic material (MacLeod & Sammel 1982; Priest et al. 1983). Regional water flow is to the NNE of the volcano (Priest et al. 1983). The volcano has many steep valleys in its flanks, but they are dry throughout the year, except for the PCR canyon. It is possible that the canyons formed during the last ice age (Russell 1905; Donnelly-Nolan et al. 2004; Donnelly-Nolan & Jensen 2009).
The climate around Newberry Volcano is strongly controlled by the orographic rain shadow of the Cascade Range (Priest et al. 1983). Lower elevations receive about 51 cm of precipitation annually (Fig. 6), and higher elevations, including the caldera, receive approximately 89 cm of rain and snow, mostly from November–April (Priest et al. 1983; Sammel & Craig 1983; Crumrine & Morgan 1994). Precipitation is probably the dominant source of recharge to EL, whereas PL has input from both precipitation and subaqueous hot springs.
Weather record for LaPine, Oregon, just west of Newberry Volcano: (a) temperature for 2010–12; (b) precipitation averages for 2010–12. Data are from: www.weatherunderground.org; accessed May 2015.
A weather record from LaPine, Oregon, just west of Newberry volcano (Fig. 6a), shows the maximum and minimum temperatures (Tmax and Tmin) for 2010, 2011 and 2012, indicating that 2011 was slightly colder than 2010 and 2012. The main precipitation events occurred in winter and 2011 was a rather dry year.
Both lakes freeze over between November and late May and are dimictic (Morgan et al. 1997). Lake levels vary with annual precipitation, but the lakes are assumed to be close to hydrologic steady state (Phillips & Van den Burgh 1968; Sammel & Craig 1983), and water levels have fluctuated only by c. 5 m since the mid-1800s, with seasonal fluctuations of less than 60 cm (Russell 1905; Phillips & van den Burgh 1968; Morgan et al. 1997). The water level of PL is controlled by a dam in PCR that has been present since the early twentieth century (Russell 1905).
Methods
Samples and field data were collected by the Wesleyan University (WU) group in August 2009, 2010, 2011 and 2012 and June 2014 and by the Central Oregon Community College (COCC) group during the 1980s and 1990s. The latter used a 20 cm diameter secchi disk attached to a plastic 30 m measuring tape to determine water transparency. A Hydrolab Minisonde and Surveyor 4a field computer were interfaced with a 100 m data cable and used to simultaneously collect depth, temperature, pH, specific conductance and dissolved oxygen (DO) data. Deep-water samples for biological and geochemical studies were obtained with a Scott 4 l sampling bottle.
The WU group measured water temperature, conductivity and DO with digital probes down to 45 m depth (YSI, models 3000-TLC and 52-DO). Water profiles were sampled with a WILCO Teflon water sampler at 10 m depth intervals down to 60 m depth. All samples were filtered with 0.2 μm nylon syringe filters in the field, and stored in 250 ml bottles without headspace. Samples for trace element analyses were preserved with a few millilitres of high-purity nitric acid. Samples for carbon isotope analyses of DIC were taken directly from the water sampler in exetainer vials pre-doped with phosphoric acid. Samples for stable isotope measurements in water were collected in small glass bottles, filled without headspace. Gases in the hot spring areas were collected through a funnel and injected into exetainer vials. Plant materials and bacterial colonies were sampled in plastic cups. Sediment cores were taken with a percussion corer with a 6.5 cm diameter plastic core barrel. Cores of 60–120 cm were extruded and sliced in 2 cm intervals and air-dried; sub-samples were then pulverized for analysis.
The water samples were analysed at WU for Cl and SO4 with a Dionex ion chromatograph (DX600), for the major cations with a Leeman inductively coupled plasma optical emission spectrometer (ICP-OES) and for alkalinity with a Mettler model 1012 autotitrator. The pH was measured in the field and in the laboratory with digital pH probes (Field: Denver model 15; lab: Fischer Accumet 90). The δD (=δ2H) and δ18O of water were measured on a Picarro instrument at the Stable Isotope Facility (SIF) at the University of California, Davis, California. Stable carbon isotope measurements of DIC and CO2 gas were performed in headspace vials at SIF using a Thermo Scientific GasBench-PreCon trace gas system interfaced to a Delta V Plus IRMS. Trace element analyses were done by inductively coupled plasma mass spectrometry (ICP-MS) at SGS (Vancouver, Canada) on the acidified water aliquots. Sediments were analysed by wavelength-dispersive X-ray fluorescence (WD-XRF; Bruker S-4 Pioneer, WU) for 25 trace elements and for major elements. All sediment samples were analysed for Hg with a Milestone DMA-80 (WU). Sediment samples were also analysed for organic carbon (Corg) and nitrogen by a CE Instruments Flash 1112 series elemental analyser (WU), and for δ13C–δ15N at SIF with a PDZ Europa ANCA-GSL elemental analyser interfaced to a PDZ Europa 20–20 isotope ratio mass spectrometer. Sediment images were collected with a JEOL JSM 6390LV scanning electron microscope (SEM) at WU. Sediment samples were analysed for 210Pb, 226Ra and 137Cs through gamma-ray counting at Yale University, Connecticut.
The Newberry crater lakes: geological observations
The floor of PL has submerged landforms, with evidence for the occurrence of multiple palaeoshorelines (e.g. Russell 1905), compositionally diverse tephra deposits, submerged hills that may represent remnants of caldera collapse and isolated boulders of possible glacial origin (Fig. 5). Bathymetric and water quality data were obtained by Phillips & van den Burgh (1968), Forcella (1982) and Johnson (1985). Reynolds (2000, 2002) used an integrated global positioning system–sonar–computer data collection system mounted on a boat to obtain spatially referenced bathymetric data and outline significant morphologic features of the lake floor.
Shoreline exposures around PL adjacent to the Little Crater campground consist of 1–2 m high wave cut cliffs of silicified blocks (Fig. 7a) composed of rounded pebbles and sand-sized pieces of pumice, scoria, lava and obsidian that are silicified into distinct 2–4 cm thick plates in which fossilized Equisetum are common. The plates form an open-tiered structure within individual blocks, with a subaerial thickness of at least 3 m. The exposure is similar to that at the Warm Springs campground, both occurring at locations of present-day hydrothermal activity. Jensen & Chitwood (2000) documented long-term uplift of the central caldera, resulting in their gradual exposure.
(a) Uplifted shoreline showing silicified plates at Little Crater campground. (b) Silicified plates and (c) tephra layers on Dante's Peak (Fig. 8) in PL. The field of view in both (b) and (c) is approximately 1 m; water depth is c. 18 m.
Submerged silicified plates (Fig. 7b, c) resemble the subaerial deposits but are thinner, and nearly ubiquitous around the perimeter of PL at depths of 12–18 m. Individual plates (Fig. 7b, c) are up to 0.5 m across and 1–2 cm thick. A submerged hill capped by silicified plates occurs near the centre of the lake (Fig. 7b, c), its top within 18 m of the surface. Here the plates consist of a silicified mix of ash and sand, the upper surface of which has embedded rounded pebbles of pumice and basaltic scoria (diameter 1–3 cm). Some plates form an open-tiered stack that is supported by silicified vertical struts separated by 1–2 cm deep cavities. Hemispherical bubble-like structures occur on the undersides of some plates.
The composite thickness of the underwater plates is less than 10 cm at most locations. Conifer branches, matted grasses and scouring rushes (Equisetum) were recovered from within the submerged plates around the perimeter of PL. Wood fragments yielded a radiocarbon age of 6.75±0.35 ka BP (Reynolds et al. 1998).
Silicification of beach sediments occurs only where there is a direct hydrothermal input. It is a near-surface process, as evidenced by the wave-rounded nature of the sand and gravel. All silicified plates thus probably originated along the shoreline close to the lake surface. The palaeoshoreline, however, differs in some characteristics from the modern one: the silicification of ancient shoreline sediment (restricted to a depth interval of 12–18 m) occurs around the entire perimeter of the lake, suggesting a past episode of significantly greater hydrothermal activity in and around the lake than is currently found. The restricted depth of the palaeoshoreline suggests a relatively rapid lake level change, with a corresponding increase in lake volume of 18–30% (Fig. 8).
North–South cross-section through PL with an interpretation of palaeoshorelines.
The age of the palaeoshoreline is constrained by the position of the plates beneath the Interlake Obsidian Flow (6.6–7.2 ka), Mazama tephra (6.85 ka BP) and the younger PL ash flow (1.3 ka BP). The radiocarbon age of 6.75±0.35 ka BP from its plant material is consistent with this stratigraphy, and the presence of the plates indicates that the two lakes were then already separated. The change in water depth may have been caused by intercaldera eruptions, earthquakes, landslides or changes in climate.
Physical limnology: long-term observations
Vertical transparency and surface temperature records have been compiled for PL since 1996 (Fig. 9). The transparency of the lake ranges from 6 to 12 m and generally increases with temperature to reach a maximum (>10 m) in August and September. The lowest recorded values (3 m) occurred in early spring. Surface temperatures range from 3°C to 20°C, with lowest values in the spring, and the highest in late August–September.
Secchi disk water transparency and surface water temperature for PL from 1996 to 2000.
The 2001 PL data show an annual pattern of thermal stratification, beginning with isothermal conditions from mid-winter through early spring (c. 3°C), followed by the development of a pronounced summer metalimnion that separates a relatively warm (12–18°C) c. 10 m thick epilimnion from a cold (4°C) hypolimnion (Fig. 10a). Waters are well oxygenated, with a pronounced DO maximum (10–13 mg l−1) at the base of the metalimnion, reaching a peak in late summer–early fall. A sample from the base of the metalimnion showed elevated levels of zooplankton (e.g. copepods) as well as phytoplankton (e.g. Volvox). The DO maximum may be caused by photosynthesizing algae in a cold, low-light ecological niche. Bottom waters have lower values of DO, in some cases down to c. 6 mg l−1 (Fig. 10b).
Monthly (January–October) water quality profiles for PL (site 041) for 2001: (a) temperature; (b) dissolved oxygen; (c) specific conductance (SpC; not compensated for temperature).
The conductivity (not temperature compensated) of PL is constant in winter, and increases with temperature in the epilimnion during the summer (Fig. 10c). The bottom waters show a strongly increased conductivity (indicating more dissolved material) without a temperature increase. More limited data for EL show a similar thermal stratification, a DO maximum in the metalimnion and much lower DO content in the hypolimnion (Fig. 11). The conductivity of EL waters is roughly half that of PL waters and again increases with temperature from early spring to summer. The pH values are constant with depth at c. 7.8 in winter, and decrease to c. 7 in the hypolimnion in summer, whereas the epilimnion acquires a higher pH (8.2) during the spring and summer.
Water quality profiles for EL in summer 2001 and early spring 2002.
The 2012–14 data also show a thermocline at 10–15 m depth in both lakes, with the hypolimnion permanently at c. 4–5°C (Fig. 12). The surface waters reached up to 20°C in August, with lower surface water temperatures in June 2014.
Water temperatures in the Newberry crater lakes in August 2011, August 2012 and June 2014 in EL (a) and PL (b), both of which show a thermocline at 10–15 m depth.
The temperature-corrected conductivities (TCCs) in PL are much larger than in EL (Fig. 13), a difference also reflected in the chemical analyses (see below). The TCC increases slightly with depth in PL, and by a greater amount in EL. Waters of EL have generally lower TCC values in June than later in the summer, and the TCC offset in EL between data from June and August is much larger in surface than in bottom waters (Fig. 13a). The TCC and temperature gradients in EL develop after ice melting and lake turnover in late May.
Temperature-corrected conductivities in the Newberry lakes in 2012 and in 2014 in (a) EL and (b) PL.
Organisms in the Newberry lakes
Primary producers in both lakes include unicellular chrysophytes and chlorophytes (Johnson 1985), as well as benthic and planktic diatoms (Fig. 14). In EL, which has a large area where light can reach the lake bottom, large benthic diatoms are common and diverse. In the shallow marginal areas both lakes have extensive beds of centimetre-sized black, green, purple or white gelatinous spheroids of Nostoc cyanobacteria, with dark green spheres dominant in PL (Fig. 15a) and white, pink and purple ones in EL (Fig. 15b). Johnson (1985) documented abundant Anabaena, another cyanobacterium, in PL. In addition, EL has floating islands (several metres long and wide) of largely subaqueous metaphytes (Fig. 15c), which may have been submerged aquatic vegetation (SAV) that became adrift.
Common diatoms from EL sediments: (a) Cymatopleura solea v. apiculata with a small, round Stephanodiscus or Dicostella on its lower-left surface; (b) Epithemia sp. with a fragment of Asterionella formosa on its upper-right surface; (c) Ellerbeckia arenaria with many small Cocconeis sp.; (d) Campylodiscus sp.; (e) Campylodiscus sp. with small Navicula sp.; (f) E. arenaria (right) covered with small Stephanodiscus or Discostella, and Campylodiscus sp. (left). Tentative diatom identifications by J. Stone.
(a) Dark green Nostoc spheroids, PL. (b) White and pink Nostoc spheroids, EL. (c) Floating SAV Island, EL. (d) Interior and exterior views of ostracod valve (Candona sp.), PL.
Both lakes have been stocked with fish since 1912 (Johnson 1985) and continue to be stocked today, but did not contain fish in 1903 (Russell 1905). Russell (1905) hypothesized that fish could not migrate up the waterfall in PCR, although crayfish did. Sediment in PL, but not in EL, contains common ostracod carapaces (Fig. 15d). Fish in EL have high mercury contents with up to c. 3 ppm Hg in large brown trout (Newell et al. 1996; Stone et al. 1996).
Chemical composition of the lake waters and hot springs
The lakes are only separated by a c. 2 km wide volcanic ridge, but differ in their chemical compositions, which are strongly impacted by the volcanic inputs (Johnson 1985). In May 2001–02, the pH throughout EL was the same, but in June and August of 2001 the bottom waters had a pH of 6.8, and surface waters values up to 8.3. In August of 2009, 2011 and 2012 the pH in EL varied from 7.5 at the surface to 6.5 in deeper waters. In June 2014, the pH ranged from 6.9 at the surface to 6.1 at depth (Fig. 16a). In PL, the pH was c. 8 at the surface and c. 7.7–7.8 at depth in August, but in June 2014 it ranged from 7 to 6.8 in one profile, and from 6.8 to 6.5 in another (Table 1). These variations in the measured pH values can be attributed in part to field temperature fluctuations and some pH measurements were done in the laboratory (at 22°C). Overall, the data suggest that in both lakes the bottom waters become more acidic after the spring melting and lake turnover, and surface waters become gradually more basic as the summer progresses, with a stronger effect in EL than in PL.
Anions in EL and PL. (a) Bicarbonate concentrations increase with depth in EL while the pH decreases. (b) c. 0.5 ppm Cl in EL with 65 ppm SO4, <5 ppm SO4 in PL with c. 3 ppm Cl. The PLHS have up to 8 ppm Cl, and the ELHS have 4 ppm Cl.
Alkalinity, pH, bicarbonate and δ13C values for dissolved carbon in lake waters and hot spring fluids
The main anion in both lakes is , with c. 130 ppm in EL and up to 400 ppm in PL. The EL alkalinity is close to that of seawater, whereas PL is three times more alkaline that seawater (Table 1). The
concentration increases with depth in EL (but not in PL), while the EL pH values decrease (2011 data, Table 1; Fig. 16a). Chloride concentrations are uniformly low, with 0.5–1 ppm Cl in EL and c. 3 ppm Cl in PL (Fig. 16b; Table 2). The EL waters have c. 65 ppm SO4 without a depth trend, whereas PL has only 2–3 ppm SO4 (Table 2; Fig. 16b).
Major element data for lake water depth profiles and hot spring samples
In the 2011 survey, the total dissolved solids (TDS) in EL was c. 5 mmol l−1 and in PL was 15 mmol l−1. In EL we find c. 25 ppm Na, 3.8 ppm K (but 5.2 ppm K in one 50 m depth sample), 26–28 ppm Ca, 12 ppm Mg and 6 ppm Si (Fig. 17a, Table 2). The Ca concentrations in the two lakes are almost identical, Mg is four times higher in PL, whereas the Na concentrations are two times higher in PL (Fig. 17b). The K concentrations are slightly higher, whereas Si is four times as high in PL (Table 2).
Cation concentrations in the two lakes in 2011, 2012 and 2014. No significant variations occur over the 2011–14 period, all profiles are constant with depth. (a) Ca and Mg concentrations in EL and PL; (b) Na and K concentrations in EL and PL.
Trace element concentrations are in the parts per billion to parts per trillion range, with most metals in the low parts per billion range, except for Cr, Mn and Zn, which occur at tens of parts per billion. PL has 14 ppb As in the water column, whereas EL has 3–4 ppb As. One 50 m depth EL sample had higher As contents as well as higher major element concentrations, close to that of PL lake water (Fig. 18). The REE concentrations in the lake waters were too low for a full assessment. The PL hot springs show a slight enrichment in Eu, suggesting the preferential dissolution of plagioclase during water–rock interactions in the underlying geothermal system (Varekamp 2015). Elemental mercury (Hg(o)) concentrations in bottom EL waters are 1.0–3.3 pM Hg(o) and 1.8 pM Hg(o) in surface waters (P. Balcom, pers. comm. 2014), which are high values compared with non-volcanic lake waters (e.g. Watras 2009). Most non-volcanic lakes have the highest Hg(o) concentrations in surface waters, whereas EL has the highest concentrations in the bottom waters.
Arsenic concentrations in PL and EL. The PL lake waters have relatively high As and no trend with depth; EL has much lower As, except for the bottom water sample that was possibly taken near a subaqueous hot spring.
The hot springs are largely mixtures of lake water and hot spring fluids, with temperatures up to 69°C at EL and 45°C at PL. These fluids have CO2 bubbles and modestly higher Cl concentrations (8–9 ppm Cl at PL hot springs), but no sulphate (Fig. 16b). The PL hot springs have up to 100 ppm Si, and substantially higher Na, alkalinity and other cations (Table 2). The EL hot springs have higher alkalinity, up to 90 ppm Si, and higher overall cation contents (Table 2).
Stable isotope data (water column)
The stable isotope composition of the lake and spring waters was measured in 2011, 2012 and 2014 (Table 3). All lake data plot on a single evaporation line in a δ18O–δD diagram, with the EL samples being the most evolved (Fig. 19). The local meteoric water line (LMWL) for Oregon was assembled from literature data (Kendall & Coplen 2001), combined with our analyses (for the Little Deschutes River, 2011; snow on the Big Obsidian Flow, 2011; Lost Lake, 2011; Lefkowitz 2012 and snow fall, 2014). Local meteoric waters (MWs) at Newberry have δ18O=−15‰ and δD=−114‰ near the intersection of the lake evaporation line and the LMWL. This value is similar to the predicted MW isotope values from regional regressions (Online Isotopes in Precipitation Calculator, Bowen & Revenaugh 2003). Over time, the isotopic compositions of the lakes varied, in a manner that broadly correlates with weather patterns (see below). Hot spring waters at PL and EL (PLHS and ELHS) have isotope ratios that are close to those of local MWs, or slightly ‘up’ the evaporation line through mixing with the lake waters. The evaporation of MWs explains the distribution of the lake water stable isotope data very well. The isotopic similarity between the hot springs and local MWs make it impossible to determine the strength of the geothermal water input using isotope mass balance calculations (e.g. Varekamp & Kreulen 2000).
Stable isotope diagram with samples from 2011, 2012 and 2014. MW values are from our samples and the literature (Kendall & Coplen 2001; squares on LMWL). The PL samples (filled circles) and EL samples (triangles) both plot on an evaporation line (2011 samples, dashed Newberry Lakes line). The 2012 and 2014 water samples are isotopically less evolved in both lakes, and plot on a slightly steeper evaporation line (thin arrow, see text).
Stable isotope data for water in EL and PL
The δ13C (DIC) values in EL (Table 1; Fig. 20) in 2014 were extremely heavy in surface waters (>+5.5‰), but slightly less so in 2011 (+4.5‰). The August 2011 profile shows a steep trend with depth (c. 4‰ gradient), whereas the June 2014 gradient was less steep (c. 3‰). Values of δ13C (DIC) in PL in 2011 and 2014 were much less extreme: 0‰ at the surface to −0.5‰ at the lake bottom (Fig. 20). The CO2 bubbles in the ELHS had δ13C=−6‰ to −8‰ (2011 data), CO2 emissions at the PLHS had δ13C=−7.4‰ (Carothers et al. 1987). These data are further discussed below in the lake carbon cycle models.
The δ13C (DIC) of PL (filled circles) and EL (triangles) in August 2011 and June 2014. There is a strong δ13C (DIC) gradient in August 2011 in EL and a weaker gradient in June 2014. Much lighter δ13C (DIC) values and much weaker δ13C (DIC) gradients were found in PL.
Sediment data
Bulk major element data
In each lake we obtained several sediment cores of c. 0.5–1.2 m length. The Corg contents differ for the two lakes: PL has 2.5–3.5% Corg, whereas EL has 7–8% Corg (Fig. 21a; Table 4). In the PL core, Corg decreases slightly with depth, but there is no simple trend in the EL core. Nitrogen concentrations in the PL core were c. 0.4% (80–30 cm depth), then increased to 0.6% and 0.9% in the surface sediment. The concentrations of N in EL are generally c. 0.8%, decreasing in the upper 10 cm to 0.4% N. The C:N (mass) values (Fig. 21b) in EL are c. 9, and the shallow and deeper sections of the PL core have C:N values of 6 and 7 respectively. A spike at 50 cm in the PL core represents a layer with much higher C:N, and slightly higher Corg.
(a) Concentrations of Corg and (b) C:N (mass) values in PL and EL sediment cores.
Sediment carbon and nitrogen concentrations and stable isotope data
Scanning electron microscope images show abundant diatom frustules (Fig. 14), and XRF analyses of the sediment show 80–90% SiO2 (Table 5a), a large proportion of which is diatom frustules. The major element XRF analyses of the sediments are not very precise, because they are outside the calibrated range of silica values (SiO2>80%). Most striking are the differences in Fe contents between sediments from the two lakes: the PL core has up to 14% Fe2O3, EL has only 1–2% Fe2O3. Other major elements contribute a few percent, but the sediments consist mainly of silica-rich material, with unusual Fe enrichments in PL. The PL sediment is also rich in P (up to 2% P2O5), containing common millimetre-sized authigenic vivianite nodules (Fe3(PO4)2 8H2O) (Fig. 22), which are identified from SEM analyses and optical microscope thin section studies. The PL core samples contain ostracod valves made of calcium carbonate (Fig. 15d).
Vivianite nodule in PL sediment.
Major element composition of EL sediment samples (core CLE12) and PL samples (core CPLA5), with two representative rock analyses from Newberry (NB) volcanics
Trace element and stable isotope ratios of lake sediment
Most trace elements in the sediment (Table 5b) are lithogenic, derived from aeolian input and volcanic ash deposition. Volcanic ash is mixed into the sediment, only occasionally present as discrete layers. Other trace elements may be derived from geothermal fluid inputs. The metals Cu, Zn and Pb occur all at low levels (several tens of parts per million), i.e. there is no evidence for anthropogenic pollution or hydrothermal enrichment (Förstner 1976). The deeper PL sediment section has Zn concentrations up to 60 ppm. The Pb concentrations are all <10 ppm Pb, well below those of polluted sediments. The Co, Ni, V and Cr concentration patterns also lack evidence for anthropogenic pollution, but V and Ni concentrations are relatively high, with up to 140 ppm Ni in PL sediment, presumably representing basaltic dust inputs. Sediments of both lakes show high Ba concentrations (PL: 250–350 ppm; EL: 100–150 ppm) that are possibly caused by the input of rhyolitic volcanic glass dust.
XRF data for major and trace elements in sediment cores of EL and PL
Most striking are the differences in the concentrations of the toxic trace elements As and Hg in PL and EL sediments. EL sediments are virtually devoid of As, whereas PL sediment shows 50 ppm As in the deepest core section, and up to 250 ppm in the middle of the core (Fig. 23a). The EL sediment has up to 4500 ppb Hg (4.5 ppm), with most samples at 500–4000 ppb Hg (Table 6a). Typical background Hg levels are 50–100 ppb (Varekamp & Buseck 1983), as found in the PL sediments (Fig. 23b). The concentrations and their trends in the cores do not suggest a regional anthropogenic origin for either Hg or As (Förstner 1976; Varekamp et al. 2003). The concentration variations downcore may point to hydrothermal input fluctuations into the two lakes over time, or diagenetic effects (Varekamp & Waibel 1987). Fish in EL have remarkably high Hg concentrations, with values up to 2.8 ppm in a specimen caught in 1994 (Table 6b). Fish in PL have very low Hg concentrations, confirming the low Hg concentrations in its sediment.
(a) Arsenic and (b) mercury concentrations in PL and EL sediment core samples.
Mercury concentrations in sediment core samples, plants and cyanobacteria from EL and PL
Mercury concentrations in fish from EL and PL
The δ13C values in sedimentary organic matter range from −22‰ to −24‰ for EL sediment (Fig. 24a, Table 4), with lighter values in PL sediment (−27‰ to −30‰). The sample with maximum C:N values in the PL core (Fig. 21b) also has a less negative δ13C value (−20‰). The δ15N values range from +0.8‰ to +1.5‰ in PL sediment and from +2 to +2.8‰ in EL sediment (Fig. 24b).
(a) Carbon and (b) nitrogen isotope ratios in PL and EL sediment, showing lighter δ13C and δ15N in PL compared with EL.
Radio isotope data of sediment cores
Core samples from EL and PL were counted for 226Ra, 210Pb and 137Cs by gamma ray counting (Table 7) and these data were used to obtain an age model for the core tops (e.g. Turekian & Bacon 2006). The results show a complex pattern with depth (Fig. 25a). In the EL core excess 210Pb (XS210Pb=210Pb-measured – 226Ra-measured) decreases irregularly with depth. The initial XS210Pb value that was used for age calculations was obtained in a log(XS210Pb) v. depth plot through linear interpolation, and ages were determined with the Constant Initial Concentration method (Robbins & Edgington 1975). The resulting age–depth plot shows three ‘age-bands’ for EL sediment (Fig. 25a), which may be the result of slumping or resedimentation, but with a mean accretion rate of c. 1.7 mm a−1. The 137Cs profile also shows an oscillating pattern, indicating reworking of the sediment with roughly the same repeats as indicated by the XS210Pb (Fig. 26b). Fewer samples were counted for radionuclides for the PL sediment core, but they indicate a broadly similar sedimentation rate. In addition to reworking of the fine-grained soft sediment, e.g. during tremors or strong lake overturns, the volcanic gas input in EL may carry 222Rn, which could locally enrich water and sediment in 210Pb and make the ages potentially more uncertain. The oscillating pattern in 137Cs, a purely anthropogenic signal, however, strongly suggests that reworking and resedimentation is a common feature, as occurs in many lakes. In conclusion, our best estimate of sediment accretion rates in both lakes is on the order of 1.5–2 mm a−1 in the top 10–15 cm. The bulk dry densities of the core materials are c. 0.2 g cm−3, giving approximate sediment mass accumulation rates of 30–40 mg cm−2 a−1.
(a) Age–depth plot for cores CLE3 (EL) and CPLA5 (PL), showing depth intervals of several centimetres in thickness with similar ages in EL, which suggests reworking/resedimentation. The few age data points for PL suggest a similar mean accretion rate of 1.5–2 mm a−1. (b) The 137Cs data from EL also show a highly variable pattern, suggestive of reworking.
Model simulations of a stable isotope–water budget algorithm for the two lakes, starting with MW. (a) The EL simulation shows good agreement between the modelled data and observations. The summer evaporation run just covers the August 2011 data and the winter mixing would move the model points back towards the MW end-member. (b) The same agreement between the model data and observations is found for PL, with one August 2011 water sample in the summer evaporation end.
Radioisotope data for sediment cores CLE3 (EL) and CLPA5 (PL)
Lake water budgets
PL is the larger of the two lakes, with an overflow through PCR, whereas EL is a small terminal lake with a small groundwater-fed lagoon on its east side (Figs 1, 2 & 5). The lake water budgets consist of precipitation inputs (both direct and through the limited watershed within the crater area), liquid geothermal inflows (mainly PL) and possible groundwater inflows. The outputs are evaporation, the PCR outflow and possible subaqueous seepage. Both lakes have small coastal subaerial hot springs with very modest water flows (Russell 1905; Forcella 1982; Johnson 1985; Ingebritsen et al. 2014) that do not contribute greatly to the lake water budgets. The water table around the lakes is low, and Newberry Volcano has no flowing rivers apart from PCR, not even during periods of snowmelt (Donnelly-Nolan & Jensen 2009). Rain and meltwater penetrate rapidly into soils and then move deeper underground. Regionally, there is snowfall in winter and modest amounts of summer precipitation (Fig. 6b). It is likely that direct precipitation on the lake surface and snow melting discharge during early spring represent the main precipitation inputs.
We evaluated the possible input of groundwater into the lakes, for which the peripheral lagoon at EL serves as an example. This lagoon water has very low dissolved constituents and has a low δ13C (DIC) value of −14‰. The lack of vertical variations in dissolved major elements in both lakes suggests that groundwater inputs are likely to be extremely small or absent, in agreement with the low water table in the caldera. Earlier water budget models used the mean water discharge of PCR (16.1×106 m3 a−1) as well as measured precipitation and evaporation rates of the lakes (Phillips & van den Burgh 1968; Sammel & Craig 1983; Johnson 1985; Carothers et al. 1987; Sammel et al. 1988; Morgan et al. 1997) to provide a first-order water budget for each lake.
Water fluxes, lake volumes, precipitation and evaporation rates from Sammel & Craig (1983) were used for stable isotope modelling of the lake waters to simulate the isotopic evolution over time. We used a 22 week summer period for evaporation and a 30 week wintertime for frozen lakes, followed by lake ice and snow melting that all enters the lake in early spring (details in Lefkowitz 2012). The model consists of weekly steps of evaporation and mean precipitation input with an outflow through PCR and a hydrothermal influx (with meteoric isotope values) for PL. The mean isotopic composition of atmospheric moisture was approximated from local mean summer temperatures (National Oceanic and Atmospheric Administration online weather data; http://www.weather.gov/climate/xmacis.php?wfo=pdt); and equation (3) in Table 8 and the mean isotopic composition of MW was obtained from the values given earlier (Table 3; Fig. 19). The mean summer relative humidity is c. 65%, which was recast (h*, Varekamp & Kreulen 2000) to account for the mean summer temperature difference between the air (9°C) and lake water (16°C) yielding h*=0.41. The equation for the slope of the evaporation line (equation (2), Table 8) also provides an estimate of the h* value (given all other values) of c. 0.4 (Fig. 19). The isotopic values of the evaporate were calculated from equation (4) (Table 8). All equations are standard expressions from textbooks on lake water cycle models (e.g. Gat 1996); please note that we use all expressions in the ‘evaporation mode’ (V/L), which means α factors are <1 and all ε values are negative; the+and−signs in all equations in Table 8 conform to the V/L usage.
Equations for stable isotope water balance model
The lake evaporation and mixing model was written in Microsoft Excel, with weekly summer evaporation cycles in the surface layer each followed by mixing of the whole epilimnion (15 m). When the lakes are frozen in winter, no evaporation or meteoric water input occurs. In the model, the winter is a single overturn event with the instantaneous addition of the cumulative winter precipitation or snow melt. The model is started with pure MW, given that the MW and hot spring water inputs have similar isotopic compositions (Table 3), and run for 40 years. With ongoing summer evaporation, the lake waters evolve along the evaporation line, to ‘fall back’ towards isotopically less-evolved (more meteoric) waters during the whole-lake late winter mixing. Both lakes reach isotopic steady state after a few decades, their maximum evolved state in isotopic space (Gat 1996).
The model results (Fig. 26) show that PL, having an outflow and subaqueous inflow, is much less evolved at steady state than the terminal EL, conforming to the analytical data. PL reached isotopic steady state close to the values observed in 2011. To obtain the most evolved EL values in 2011, late summer model values agree with the analytical data from the August 2011 EL samples. In 2012 and 2014, the lake waters were isotopically less evolved, because these years were wetter and slightly warmer than 2011 (Fig. 6).
The agreement between the modelled and observed isotope values using the Sammel & Craig (1983) water budget is quite good (Fig. 26; Table 9). For EL, the water budget model is close to steady state, and some minor seepage may occur. For PL, a large input deficit appears in the older model (Table 9), suggesting a much larger hydrothermal input to achieve closure. Evidence for subaqueous water input includes: (1) ‘hot zones’ in PL (on the NW side of lake) that create thin ice or ice-free areas during the winter (Phillips & van den Burgh 1968; Forcella 1982; Johnson 1985; Morgan et al. 1997), and (2) the chemical hot spring components in the sediment (see below) and the export rate of chemicals through PCR, which attest to a substantial inflow of hot spring waters into PL. The earlier estimate of the hydrothermal input (Sammel & Craig 1983; Morgan et al. 1997) creates a large closure error in the water balance (Table 9). Below, we constrain the magnitude of the hydrothermal inflow by combining the stable isotope model with chemical mass balance.
Water budget model of Sammel & Craig (1983)
Sediment chemistry, hot spring inputs and lake mixing
The lake sediments consist largely of diatom remains and organic matter with some volcanic components, hereafter called ‘ash’, although it may be a combination of primary pyroclastic material (ash fall-out from eruptions) and aeolian inputs from the surrounding bare pumice and cinder fields. A comparison between a Newberry rhyolite, basalt and EL sediment (Table 5a) suggests that most of the sediment consists of biogenic silica (BSi, diatom frustules) mixed with some ash. Trace element systematics can be used to estimate the ash contributions. We used the Zr and Y concentrations (Table 5b; Fig. 27) to calculate the contribution of ash to the lake sediment. The Zr/Y values in many volcanic ashes at Newberry range from 6 to 8 (Kuehn & Foit 2000). The sediment data for both lakes show a slope close to this value (Fig. 27), and we calculate that volcanic ash may account for 10–30% of the sediment in both lakes.
Proportions of volcanic ash in lake sediment samples. The Zr–Y relation in lake sediment is close to that in Newberry Volcano ashes (Kuehn & Foit 2000; Kuehn 2002). The data suggest that the core sediments contain 10–30% volcanic ash.
The main geothermal inputs in EL are CO2 and H2S, with traces of Hg, and a small associated input of hot water with dissolved constituents. The PL sediments have high concentrations of Fe, Mn, P and As, interpreted to be hydrothermal components. The hot spring waters have high concentrations of the major cations (Na+, K+, Mg2+, Ca2+; Table 2) and presumably similar fluids are also vented from the submerged springs in PL. Those four major elements remain largely in the water, whereas Fe2+, Mn2+, and arsenite or arsenate partition largely into the sediment. The hot fluids are presumably also rich in dissolved silica, so part of the silica in PL sediment may be of hydrothermal origin. The dissolved major element concentrations in PL have not varied much over the sample years (1964, 1980, 2010, 2011, 2012, 2014), so we assume that PL is broadly in chemical steady state. The dissolved silica in PL is brought in by the hydrothermal fluids, and lost through burial of diatom frustules, direct precipitation through cooling/mixing of hot water with lake water and export through PCR. The lake has c. 20 ppm Si (Table 2), which may be close to steady state as well. The hydrothermal inputs in PL must be considerable in magnitude because of the high concentrations of Fe, P (in vivianite; Fig. 22) and As (Fig. 23a) in the PL sediment.
A first-order approach to estimate the subaqueous hot spring input assumes that the export of conservative chemicals through PCR equals the input from the hot springs. The PCR waters annually export c. 300 t of Si (calculated from the PCR mean water flux multiplied by the Si contents of PL water). The Si sequestration rate into the sediment (diatoms+direct silica precipitation) at linear sedimentation rate of 2 mm a−1, a bulk dry density of 0.2 g cm−3, 15% ash in the sediment and 80% SiO2 in the bulk sediment equals c. 1000 t a−1 of Si. The total annual Si loss term (Si burial in sediment and PCR export) is c. 1300 t a−1. We assume that the submerged lake bottom hot springs have to bring in a similar amount of Si to provide the near constancy of dissolved silica over the years.
The major dissolved cations in the hot spring fluids reflect the chemical and thermodynamic controls of the underlying geothermal system on the hot water composition (e.g. Truesdell 1984). Drill hole data suggest maximum geothermal reservoir temperatures of 265°C (Keith & Bargar 1988; U.S. Bureau of Land Management 2007), whereas lake water geothermometry using the Na/K–temperature relationship (Fournier & Truesdell 1973; Cann 2002) suggests an original temperature of c. 235°C. Presumably, water in the subaqueous hot springs stems directly from the high-temperature reservoir. Dilution, boiling or conductive cooling do not change Na/K values markedly, so hot springs may retain their original Na/K values at much lower temperatures. The silica concentration in the hot spring waters, however, is largely a function of water temperature. We equate the silica export rate through PCR and the Si sequestration rate into the sediment with the hydrothermal Si input rate, and the temperature of the springs then determines the hydrothermal water input. At 235°C, the fluids could carry c. 200 ppm Si (Verma 2000), leading to a required hydrothermal water influx of 6.5×106 m3 a−1, insufficient to remove the closing error in the water balance (Table 9). If the temperature of the hot springs was slightly lower (c. 135°C with c. 100 ppm Si), this would close the gap in our model water balance (12×106 m3 a−1 hot spring influx instead of the 3×106 m3 a−1; Table 9). Therefore, as a first approximation, we assume that PLHS have a temperature of c. 135°C and a water flux of 12×106 m3 a−1. From the PCR export rate for conservative elements a total element cycling budget can be drawn up, broadly defining the composition of the hot spring fluids. This is an ongoing effort with more analytical data being collected at depth in the thermal zone of PL, with the ultimate goal of a combined stable isotope, physical mixing, and chemical mass balance model for PL. Currently, we estimate a water residence time of 8 years for PL and c. 20 years for EL.
The hot spring input could impact the water temperatures, warming the bottom waters from c. 4°C after the spring lake turnover to 5–6°C later in the year, as some of our recent water temperature profiles suggest. The bottom water warming can also be related to slow mixing through wind stresses, bringing warm water from the epilimnion down to greater depth. Basin morphology and the prevailing wind direction affect vertical water mixing in lakes (Wetzel 2001), and the lack of vertical gradients in the stable isotopes of the water and cation contents suggests that a slow mixing process may be involved in lake turnover during the summer, even though the water column is thermally stratified (Figs 10, 11, 12). Throughout the year, winds in the caldera are mainly from the west, so that water piles up at the east shore in both lakes (Fig. 28). With little shoal area (Fig. 5), PL waters pile up on the eastern shore, displacing water downwards and resulting in deep vertical mixing early in the season. This circulation is possibly made more vigorous by the westward outflow through PCR.
Simple mixing profiles for PL (left) and EL (right), with whole-lake mixing in PL and more compartmentalized mixing in EL.
In contrast, much of the EL basin consists of shoals (Figs 4 & 5). Water in the shoals tends to remain there, warms up, and forms a separate, warm, low-density layer. Surface water in the deep central basin is blown eastwards, but not replaced by vertical mixing. The wind-driven water sloshes around the shoal area, resulting in less complete, early-season vertical mixing. We suggest that the timescale of mixing in PL is slow compared with the timescale of ambient heating of the epilimnion, creating a thermally stratified lake that slowly mixes. Given the profound compositional and isotopic differences between EL and PL, seepage down the hydraulic gradient from EL to PL is probably insignificant, contrary to the suggestion by Sammel & Craig (1983).
The lake carbon cycle
The concentrations of DIC in water and Corg in sediment, combined with their δ13C (DIC) values, provide insight into the carbon cycles of the two lakes. The lakes have small watersheds that are largely covered with pine trees (Russell 1905; Johnson 1985). Little debris from these evergreens is carried into the lakes (with the exception of pollen grains), so the sedimentary organic matter is likely to be produced largely in situ. The burial of primary producers and heterotropic aquatic organisms contribute to buried Corg, with some buried SAV remains (Fig. 29). The C:N (wt) values in PL sediment are 6–8 and c. 9 in EL sediment, both pointing to freshwater algae as the main contributors (Meyers & Teranes 2001). The δ15N and δ13C values in the sedimentary organic matter suggest that PL sediment contains a mixture of cyanobacterial matter (including the Nostoc balls, Fig. 15a, b) and diatom frustules (Fig. 14), whereas in EL the SAV debris from the floating plant islands is present (Figs 15c & 29).
Composition of organic matter in the lake sediments. Cyanobacteria and diatoms are found in PL, with additional SAV debris (macrophytes) found in EL. The stable isotope mixing relations provide estimates of the contributions of each end member.
The concentration increases slightly with depth in EL while pH decreases, probably the direct result of the absorption of geothermal CO2 bubbles at its lake bottom. No depth trend in
concentration or pH is found in PL, which may be explained by the absence of direct CO2 gas inputs and/or better overall mixing. The δ13C (DIC) depth gradient in EL (for August 2011) is very steep, from +4.5 ‰ at the surface to near 0‰ at the bottom, whereas the δ13C gradient is much weaker in PL at that time (Fig. 20). The DIC can be calculated from the alkalinity, pH and water temperature through use of a speciation program (webphreeq; https://www.ndsu.edu/webphreeq/). From the measured δ13C (DIC), the calculated quantities of carbonate species and the fractionation of intercarbonate species carbon isotopes (Clark & Fritz 1997) we obtain a value of δ13C (CO2(aq))=−3.5‰ to −7‰ for the summer–winter temperature range using isotopic mass balance calculations. If the δ13C (organic matter) in EL is c. −23‰ and the algae use CO2(aq), the δ13C photosynthetic offset would be c. 20‰, close to accepted values (e.g. Clark & Fritz 1997; Finlay 2004).
The δ13C isotopic gradient in EL is partially explained by epilimnial photosynthesis, which preferentially takes up isotopically light carbon (δ13C c. −23‰), combined with sluggish vertical mixing. The absence of a δ13C (DIC) gradient in PL may relate to its more active mixing regime. The organic matter sinks to the lake bottom, where part of it is buried (up to 8% Corg in EL sediment; Fig. 21a) and part is oxidized to CO2 through respiration or other oxidation pathways, providing the O2 depletion in deeper EL waters (Fig. 11). In addition, O2 might be consumed through oxidation of H2S to sulphate. The gas bubbles in the ELHS and PLHS have a δ13C value of −7‰ to −8‰, although values of −3.5‰ were measured by us in recent years. If the bubbles rising from the EL lake bottom have similar δ13C values, and are completely dissolved (no net fractionation), the DIC in the whole lake has become isotopically heavier over time through these repeated cycles of photosynthetic fractionation and burial of isotopically light Corg.
Another process in the EL surface waters is diffusive CO2 loss, because the calculated partial pressure of CO2 (PCO2) in EL is well above the ambient PCO2 value (Pérez et al. 2011; Figs 30 & 31). If CO2 gas escaping from EL has a δ13C value of c. −3.5‰, this process will leave an isotopically heavier residue. During summer, combined photosynthesis and diffusive CO2 loss in the epilimnion will thus create a DIC pool with isotopically heavier carbon, creating the steep vertical δ13C(DIC) gradient in EL (Fig. 20). The δ13C gradient becomes steeper from spring to fall (June 2014 v. August 2011, Fig. 20) and the whole lake DIC pool may become isotopically heavier over the years. The CO2 input at the bottom of EL thus can be constrained from the isotopic mass balance of the diffusive CO2 loss and burial rate of photosynthetic material (Capece et al. 2016).
Carbon cycles in EL (top) and PL (bottom). EL, with its strong δ13C (DIC) gradient, has abundant diffusive CO2 release with additional photosynthetic CO2 withdrawal in the epilimnion. The calculated δ13C of dissolved CO2 is indicated at the top of both diagrams. The input δ13C values were measured in bubbles in the hot springs, and the δ13C (DIC) for the hot spring input in PL was based on equilibrium with bubble δ13C. The water residence times (RT) were approximated from the water budgets given in the text.
PCO2 in the Newberry lake waters calculated with the Phreeq programme. EL shows strong oversaturation with respect to atmospheric PCO2 (currently 400 ppmv CO2), with a more modest oversaturation for PL.
The carbon cycle processes in PL are comparable to those in EL (Fig. 30), but the aqueous PCO2 is much lower, so that the diffusive CO2 loss also should be much smaller. The summer δ13C value of CO2(aq)=−8.5‰ and that of organic matter in PL sediment is c. −28‰, creating an offset of 20‰ between dissolved CO2 and photosynthate, similar to EL. The carbon export through PCR does not involve isotopic fractionation, so the small δ13C isotopic gradient with depth in PL is likely to be largely due to photosynthesis. The PL sediment has about half of the Corg (4%) of EL sediment (8%) at roughly equal sediment mass accumulation rates, and we conclude that primary productivity is lower in PL than in EL (Johnson 1985). The PLHS contribute dissolved carbonate, possibly with δ13C at 0‰ in isotopic equilibrium with the CO2 gas at the PL hot springs. This would imply that the carbon isotopic evolution of PL waters is largely driven by the burial of Corg, which is a relatively small term, in agreement with the mean δ13C (DIC) in PL, which is almost equal to the presumed δ13C value of the hot spring input.
Conclusions
The two Newberry crater lakes were once part of a larger caldera lake system, but became separated c. 7 ka BP. Since that time, the two lakes have had very different compositions and water budgets. Drowned palaeoshorelines (as indicated by silicified beach materials) are now c. 18 m below the current surface of PL, and formed when volcanic inputs were probably more voluminous than today. Both lakes belong to the carbonate-rich volcanic lakes clan (Varekamp 2015; Varekamp et al. 2000), but are profoundly different in chemical composition with regards to major elements, alkalinity and trace elements; PL is approximately three times more concentrated than EL. Sediments in both lakes consist largely of diatom frustules (BSi), volcanic ash and organic matter, with a significant hydrothermal component in PL as well. Existing water budgets for the two lakes were evaluated with models of water stable isotope evolution, and suggest that both lakes are close to isotopic steady state. However, the hydrothermal input component into PL must be about four times larger than has previously been suggested. The water residence time is c. 8 years in PL and 20 years in EL.
EL has a (visible) input of CO2-rich bubbles that dissolve before they reach the surface (e.g. Caudron et al. 2012), which causes the slightly lower pH values in EL compared with PL, and the strong difference in internal PCO2. The CO2 bubbles probably also carry H2S, which is oxidized to SO4 in the lake water, and elemental Hg vapour (Varekamp & Buseck 1984). The EL waters thus have elevated dissolved Hg, its sediments are Hg-rich (up to 4.4 ppm Hg), and its large fish have up to 2.8 ppm Hg, an ecological threat for humans and animals (Suchanek et al. 2008). The conversion of Hg(o) to methylated Hg may be mediated by sulphate-reducing bacteria (Compeau & Bartha 1985), which would thrive at the high sulphate contents in EL. The dissolved silica sink from the abundant diatom productivity in EL is probably balanced by a small input of hot spring waters.
In contrast with EL, PL has major hydrothermal fluid inputs at its bottom, which leads to high alkalinity with bicarbonate concentrations up to 400 ppm. PL sediments have a hydrothermally impacted chemical composition, with high As and Fe concentrations and elevated Mn and P levels. The hot springs in PL supply the main dissolved cations to the lake waters as well as Fe, P, As and Mn to the sediment. The Fe is probably largely sequestered as iron oxides in the sediment (up to 14% Fe2O3), which may adsorb the As (up to 250 ppm in the bulk sediment). Some As may be associated with the diagenetic mineral vivianite, replacing P in its structure. The As concentrations in the PL waters (10–15 ppb As) are close to the drinking water limit for As (WHO 1996); elevated As is typical for many volcanic thermal springs (Lopez et al. 2012). The EL waters and sediment have very low As concentrations, whereas the PL sediment has background Hg concentrations. The hydrothermal Si becomes largely sequestered into the sediment as diatom frustules, although part may be directly precipitated in the PL sediment as hydrothermal silica during cooling of the hot input fluids. The various data suggest more active vertical mixing in PL than in EL, partially driven by winds in addition to resupply by the subaqueous hot spring waters and surface overflow through PCR.
The lake ecosystems thrive on the geothermal input of CO2, silica and P. The fixed nitrogen is supplied by cyanobacteria, including the Nostoc balls and possibly other cyanobacteria. EL has the unusual floating islands of plants (possibly disturbed SAV), which, with the diatoms, other unicellular eukaryote algae and cyanobacteria make up the bulk organic matter in EL sediment. PL does not have an SAV component, so the organic matter consists of diatom remains, other eukaryote unicellular algae and cyanobacteria. We have not yet studied the lacustrine ecosystems in detail, but a major shift may have occurred in PL less than 100 years ago, possibly related to the introduction of fish, as indicated by the sudden shift in C:N values around that time.
Steep vertical gradients in δ13C (DIC) are present in EL, presumably related to photosynthetic CO2 withdrawal and CO2 diffusive losses at the lake surface. With its high local PCO2, EL might have been an equivalent of Lake Nyos in the past, when the climate was colder and annual overturn possibly did not occur. The floods through the PCR canyon (Chitwood & Jensen 2000) may have been caused by dam failure, volcanic activity or possibly CO2 oversaturation and bubble evolution processes. Research to evaluate the volcanic and limnological hazards and possibly develop parameters for volcano monitoring is ongoing.
We speculate that geothermal fluids rise below the lakes from the east side (Fig. 32). When they reach a vertical structure related to the emplacement of the volcanic ridge that separates the two lakes, the fluids become sufficiently shallow to develop a separate gas phase (CO2, H2S and traces of Hg). These gases enter EL, acidifying the lake and causing enrichment in CO2, sulphate and Hg. The residual fluids, rich in Fe2+, Si, carbonate, P, Mn and As, find their way further up west and enter PL. This leads to the characteristic PL Fe–As-rich sediments and its relatively concentrated waters.
Conceptual model of the two Newberry Volcano crater lakes. Rising geothermal fluids undergo phase separation below EL, with gas entering EL and the residual aqueous fluid entering PL.
The two Newberry Volcano lakes are unique in their chemistries and gas contents, and have few analogues, although several Italian volcanic CO2-rich lakes may have strong similarities (Carapezza et al. 2008; Chiodini et al. 2012; Cabassi et al. 2013). At Newberry Volcano and the Italian lakes, sudden intense CO2 degassing is a possibility through seismic triggers or disrupted lake stratification, and presents a volcanic hazard that must be considered (Evans et al. 1993; Caudron et al. 2012; Chiodini et al. 2012).
Acknowledgments
The research was largely based on the undergraduate thesis work of J. N. Lefkowitz and was funded by the Wesleyan University Harold Stearns Chair Fund to J. C. Varekamp. J. N. Lefkowitz was funded by the McKenna Fund and a Mellon grant from Wesleyan University. We appreciated help in the field by James Rae, Dylan Varekamp, Scott Herman, and Al Waibel and his family. Sarah Lerman-Sinkoff took the SEM photographs, and Dr. Jeffery Stone (Indiana University, Terre Haute) identified the diatoms. The East Lake Resort owners kindly provided logistical help that is acknowledged. The XRF analytical facility at Wesleyan University is funded by the NSF under grant NSF EAR-0949376.
- © 2017 The Author(s). Published by The Geological Society of London. All rights reserved